极值分布基于最佳线性无偏估计的异常数据检验

Outlier Test for Extreme Value Distribution Based on the Best Linear Unbiased Estimation

文章针对样本数据服从极值分布、样本中出现异常数据的检验问题进行了研究。首先基于尺度参数σ的最佳线性无偏估计(BLUE),并利用次序统计量的贡献率,构造了检验统计量,然后通过Monte-Carlo模拟得到了检验统计量的分位数,给出了异常数据的疑似个数和具体检验方法,最后通过实例说明所给出的检验方法是可行的。

This paper studies the problem that the sample data obeys the extreme value distribution and abnormal data appears in the sample.Firstly,based on the best linear unbiased estimation(BLUE)of scale parameters,the paper adopts the contribution rate of order statistics to construct the test statistics,and then uses Monte Carlo simulation to obtain the quantile of the test statistics,and presents the suspected number of abnormal data and the specific test method.Finally,the paper employs an example to illustrate the feasibility of the proposed test method.